We taught a collar to notice when a dog isn't well.

Oggii turned a coin-sized sensor into daily clinical alerts for vets and pet owners. Our algorithms ran inside Zoetis's PetDialog+ and Elanco's platform, on 100,000 devices.

Oggii operated from 2013 to 2021. Its IP was acquired by Elanco.

Clinical alert: significant increase in daily scratching
Live behaviourSheba, Siberian Husky, 3
Walking
confidence 94%
Daily scratching, rolling 3-day average
0devices sold through Zoetis and Elanco
0tagged behaviour patterns fed to the models
0dog days tagged by vets
0accuracy, independently validated

In a nutshell

  • 1

    Illness detection

    Pain, arthritis, atopic dermatitis and otitis, spotted from behaviour.

  • 2

    Personalisation

    Food, medical treatments and over-the-counter products matched to each dog.

  • 3

    Device agnostic

    An algorithm that could serve any smart device on the market.

A young Australian Shepherd looking up

The algorithms, level 1 (2013–2018)

Ten behaviours, recognised second by second from a collar.

EatingScratchingDrinkingSleepingOff collarRunningWalkingSprintingStaticShaking
0dogs videotaped, tagged and labelled wearing Oggii
0videotaped, tagged and labelled patterns
0tagged behaviour patterns fed to machine learning
  1. Preprocessing

    Decompression, sampling-rate unification, axis alignment, SMOTE, 2–4–8–12 second moving windows.

  2. Feature selection

    From 100+ features down to the best set for each target behaviour.

  3. Training

    Based on a pre-balanced matrix score for the best success rate.

  4. Machine learning

    A random forest of 30 decision trees, 17 levels deep, voting by majority.

  5. Behaviour models

    The most probable path of a dog's behaviour, time-of-day dependent, with enhancement models.

  6. Result

    Behaviour identification at 90% average accuracy: 24-hour values, night values, activities.

Validated by the University of Surrey

An independent team at vHive, the university's veterinary health innovation engine, checked Oggii against tagged video, second by second. Accuracy came in above 90%.

Read the paper in PLOS ONE

Level 2: Vet AI (2017 onward)

A dataset combining 50 years of dog data: what owners saw every day, what vets recorded, and what the collar measured.

Owners' daily data

  • Diet, snacks and medication times
  • Walks, baths and grooming
  • House guests, boarding, changes at home
  • Vet visits and special activities
  • Owner changes
  • The owner's impression of the dog's day

Medical records

  • Full records, at least a year back
  • Interview with the attending vet
  • Monthly vet check-ups
  • Unplanned vet visits

Alert and state algorithms

  • 1, 3 and 7-day deviations
  • 2 to 60-day trends
  • 24-hour values
  • Night values
  • Activities

20,000+ dog days, tagged by several vets

Each day was labelled for background disease, the presence and type of symptoms, their severity and the vet's certainty. Vet AI learned from those days to identify, rate and report pain, arthritis, atopic dermatitis and otitis.

Severity1 to 9
Certainty1 to 4
Symptomspresent, and which
Day #10,409
Background disease: none
Symptom: present
Type: otitis
Severity: 8
Certainty: 1

Clinical value, every day

Oggii produced a daily, scored and categorised status for each dog, translated into plain-language messages for the vet and the owner.

General health alerts
Presence and severity of symptoms
Significant health trends
Phone app showing an ear infection alert
Zoetis PetDialog vet dashboard listing clinical alerts

Clinical alerts in the Zoetis PetDialog clinic dashboard, powered by Oggii.

The Oggii 3 sensor, 1.8 by 0.8 by 0.5 inches

Oggii 3. For dogs. For cats.

  • Rechargeable line90 days between charges
  • Sealed lineone-year battery
  • Setupconfigurable from the app
  • Buildbite-proof, waterproof to IPX7, recyclable
  • CertificationCE and FCC
  • Size1.8 × 0.8 × 0.5 inches
  • Manufacturing cost$25

The business

Zoetis
Commercial and R&D partnership
  • Strategic partnership since 2015
  • 32,500 devices sold
  • Soft launch in four territories
  • R&D contract
  • Service and support contracts
Elanco
Commercial partnership, then acquirer
  • Oggii as the backend
  • Algorithm licensing
  • Serving 5,000 devices from other vendors
  • Acquired Oggii's IP in 2021

Zoetis launch results

0%

of monitored dogs became clinical cases within 60 days

Dogs that looked fine to their owners were brought in because the data showed otherwise.

What the alerts found

  • Dermatologic 43%
  • Post-op 24%
  • Pain 24%
  • Orthopedic 5%
  • Anxiety 5%
0%more vet visits
0%more medication
Before
With Oggii

3×market share in Oggii-initiated prescriptions

All of this together made

a commercial tool clinics actually used

Selected use cases

Sixteen real cases from clinics in South Africa, New Zealand and Belgium. Open any dog to see its data.

The company

  1. 2013Founded in Tel Aviv with the 8200 EISP accelerator, to decode dog behaviour from movement.
  2. 2015Strategic commercial and R&D partnership with Zoetis. Oggii's algorithms power PetDialog+.
  3. 2017Independent validation by the University of Surrey, published in PLOS ONE. Level 2 and Vet AI begin.
  4. 2018Zoetis soft launch in four territories.
  5. ElancoAlgorithm licensing, with Oggii as the backend for devices from other vendors.
  6. 2021Oggii's IP acquired by Elanco.